Got

Like Git, but with an 'o'

Got is version control, with a model similar to Git's. A solution to the problems that come from using Git to store all of your data. Got uses an improved data structure that better handles large files and directories, and encrypts all the data that you give it.

Quick Links

• ARCHITECTURE.md
• CLI Docs
• Cryptography Overview
• GotFS
• GotKV

Features/Roadmap

• Content-Defined chunking of large files into blobs with a maximum size.
• E2E encryption. Branch names, directory names, file names, and file sizes can all be hidden from remote repositories.
• Efficiently add/remove large files and directories to/from existing filesystems.
• Stage changes with add, rm, put, and discard commands.
• Inspect branch state with cat and ls.
• Print and change the active branch with active command.
• Commit changes, with commit.
• Create, delete, and list branches with branch commands.
• Copy one branch state to another with sync.
• Share repositories using the Blobcache protocol with serve command.
• Efficiently pack many small files into fewer blobs.
• Checkout the head of a branch to the working directory.
• Merge 2 branches.

Getting Started

Installation

Either download a prebuilt binary or build one from source.

Installs to $GOPATH/bin with just. If that isn't on your path, just copy the executable from there to wherever you want.

$ just install

Create a New Repo

Then initialize a repository in the current working directory. Make sure you cd to where you want the repository.

$ got init

That will create a new repo using an in-process blobcache. All content will be stored in the .got/blobcache directory.

This is the recommended way to try out blobcache.

A repo can also be initially configured to use the system (or any) blobcache instance. There are two ways to do this:

1. Using BCP with got init --blobcache-remote <bcp endpoint>
2. Using HTTP with got init --blobcache-http <http endpoint>

For large repositories, it is recommended to use an out of process blobcache

The HTTP endpoint should just work, the BCP endpoint requires granting access to Got.

Contributing

We use just as a command runner. All of the common development tasks have just commands.

To run the tests:

$ just test

To build release binaries

TAG=v0.0.x just build

Where TAG is an environment variable which should be set to the release version, or the Git hash of the source used for the build. The release binaries will be under the out directory.

Got in Action

Got importing a 3GB file.

$ got add large_file.dat
[1.608s]
  [1.564s] large_file.dat data_in=(1.24GB Δ=799.35MB/s)
    [1.564s] worker-0 blobs_in=(182blobs Δ=117.26blobs/s), data_in=(157.09MB Δ=101.21MB/s)
    [1.564s] worker-1 blobs_in=(197blobs Δ=127.43blobs/s), data_in=(158.19MB Δ=102.33MB/s)
    [1.564s] worker-2 blobs_in=(181blobs Δ=116.50blobs/s), data_in=(154.83MB Δ=99.66MB/s)
    [1.564s] worker-3 blobs_in=(172blobs Δ=112.63blobs/s), data_in=(154.01MB Δ=100.85MB/s)
    [1.564s] worker-4 blobs_in=(179blobs Δ=116.82blobs/s), data_in=(157.34MB Δ=102.68MB/s)
    [1.564s] worker-5 blobs_in=(177blobs Δ=115.36blobs/s), data_in=(156.12MB Δ=101.75MB/s)
    [1.564s] worker-6 blobs_in=(199blobs Δ=130.09blobs/s), data_in=(155.20MB Δ=101.46MB/s)
    [1.564s] worker-7 blobs_in=(161blobs Δ=104.50blobs/s), data_in=(150.69MB Δ=97.81MB/s)

More

Read more about the configuration objects in doc/Config.md.

Support and development discussion happen in the INET256 discord.